The location of the axon initial segment affects the bandwidth of spike initiation dynamics

Christophe Verbist, Michael G. Müller, Huibert D. Mansvelder, Robert Legenstein, Michele GiuglianoPLOS Computational Biology 2020 · DOI · OpenReview

Introduction

In neuroscience, neurons are frequently conceptualized as simple integrators of synaptic inputs. However, they are sophisticated filters whose signal-processing capabilities are deeply tied to their physical structure. A critical component of this structure is the Axon Initial Segment (AIS), the specialized region where action potentials (APs) are initiated. The location of the AIS is not fixed; it varies across cell types and can shift position during development or in response to activity—a phenomenon known as AIS plasticity. This study investigates the functional consequences of AIS geometry, specifically how the distance between the soma (cell body) and the AIS affects a neuron's "bandwidth"—its capacity to follow rapidly fluctuating inputs.

Methodology

To bridge the gap between single-cell geometry and network-level computational power, the researchers employed a multi-scale modeling approach using the NEURON simulator. The study progressed from simplified ball-and-stick (BAS) models to biophysically detailed 3D reconstructions, including 65 types of rat cortical neurons and a human pyramidal neuron model.

Methodological framework for assessing neuronal bandwidth and the impact of AIS location.

The core of the analysis involved Linear Transfer Gain Analysis. By injecting "noisy" sinusoidal currents into the models and measuring the output firing rate across a range of frequencies (1 to 1000 Hz), the authors calculated the Cut-Off Frequency (COF). This metric defines the upper limit of the neuron's operational bandwidth. Additionally, the researchers used Liquid State Machines (LSM) to evaluate how these cellular-level changes impact the performance of a neural network in solving non-linear classification tasks, such as the delayed XOR problem.

The Biophysics of Spike Initiation

The study's findings are rooted in "resistive coupling theory," which posits that the soma acts as a massive electrical load (or sink). When the AIS is located close to the soma, this electrical load "blunts" the spike initiation process. As the AIS moves further away, it becomes "decoupled" from the soma's load, allowing for a significantly sharper and faster spike onset.

Comparison of spike initiation dynamics between proximal and distal AIS locations. The phase plot (right) shows a sharper onset for the distal AIS.

This increased sharpness is the key to bandwidth expansion. A sharper spike onset allows the neuron to track the high-frequency components of its input more accurately. The authors observed that increasing the AIS distance can boost a neuron's bandwidth by up to 150% in specific cell types.

Key Insights and Network-Level Impact

The sensitivity to AIS location varies significantly across the cortex. For instance, Layer 2/3 pyramidal cells exhibited extreme sensitivity, whereas Layer 4 cells were more robust. In rat thick-tufted pyramidal cell (TTPC) models, moving the AIS increased the COF by 55% (from ~275 to ~450 Hz). Human pyramidal neurons showed a more modest ~20% increase, likely due to their larger dendritic trees acting as a dominant electrical load.

Beyond individual neurons, these structural changes have profound implications for network performance. In Liquid State Machine simulations, shifting the AIS to a more distal position improved the network's accuracy in the delayed XOR task from approximately 86% to over 90%. Interestingly, these performance gains showed a "saturating" effect, with the most significant improvements occurring within the first 20 μm of displacement.

Conclusion

This research demonstrates that the physical location of the AIS serves as a structural tuning knob for neural signal processing. By modulating its bandwidth, a neuron can effectively adjust its contribution to network-level computations. While the study focused on excitatory neurons and utilized point-neuron simplifications for network tasks, it provides a compelling framework for understanding how AIS plasticity might regulate the computational capacity of the brain. Future work will need to explore these dynamics in inhibitory interneurons and seek direct experimental validation of bandwidth shifts in living circuits.

Summary updated

Abstract

Author summary The neurons in the brain encode information through electrical impulses. The performance of a cell in terms of its ability to process and transfer information downstream thus depends heavily on the machinery of initiation of these impulses. In this work, we consider both the cell morphology and the biophysical properties of impulse initiation as the primary parameters that influence information processing in single neurons, as well as in networks. We specifically analyze the location of nerve impulse initiation along the cell’s axon and the way the neuron transfers incoming information. By using single-cell models of various complexity as well as network models, we conclude that information processing is sensitive to the geometrical details of impulse initiation.
BibTeX
@article{verbist2020location,
  author = {Christophe Verbist and Michael G. Müller and Huibert D. Mansvelder and Robert Legenstein and Michele Giugliano},
  title = {{The location of the axon initial segment affects the bandwidth of spike initiation dynamics}},
  journal = {PLOS Computational Biology},
  year = {2020},
  doi = {10.1371/journal.pcbi.1008087},
}
Figure